Use of Highly Variable Intergenic Spacer Sequences for Multispacer Typing of Rickettsia conorii Strains

Abstract

By use of the nearly perfectly colinear genomes of Rickettsia conorii and Rickettsia prowazekii, we compared the usefulness of three types of sequences for typing of R. conorii isolates: (i) 5 variable coding genes comprising the 16S ribosomal DNA, gltA, ompB, and sca4 (gene D) genes, which are present in both genomes, and the ompA gene, which is degraded in R. prowazekii; (ii) 28 genes degraded in R. conorii but intact in R. prowazekii, including 23 split and 5 remnant genes; and (iii) 27 conserved and 25 variable intergenic spacers. The 4 conserved and 23 split genes as well as the 27 conserved intergenic spacers each had identical sequences in 34 human and 5 tick isolates of R. conorii. Analysis of the ompA sequences identified three genotypes of R. conorii. The variable intergenic spacers were significantly more variable than conserved genes, split genes, remnant genes, and conserved spacers (P < 10⁻² in all cases). Four of the variable intergenic spacers (dksA-xerC, mppA-purC, rpmE-tRNA^fMet, and tRNA^Gly-tRNA^Tyr) had highly variable sequences; when they were combined for typing, multispacer typing (MST) identified 27 different genotypes in the 39 R. conorii isolates. Two batches from the same R. conorii strain, Malish (Seven), with different culture passage histories were found to exhibit the same MST type. MST was more discriminatory for strain genotyping than multiple gene sequencing (P < 10⁻²). Phylogenetic analysis based on MST sequences was concordant with the geographic origins of R. conorii isolates. Our study supports the usefulness of MST for strain genotyping. This tool may be useful for tracing a strain and identifying its source during outbreaks, including those resulting from bioterrorism.

The growing challenge presented by strains of bacterial pathogens with increased virulence and/or transmissibility, strains with antibiotic resistance, and strains used in bioterrorist attacks has highlighted the requirement for effective methods to identify such strains and track their spread (26). Key factors in the control of these strains are rapid detection, appropriate therapy, and contact tracing to arrest further transmission. To date, molecular characterization of bacterial strains has been pursued for two different objectives (57, 60). The first is global or long-term epidemiology studies that monitor how microbial populations change over time. Much of the current knowledge of bacterial population genetics (38, 52, 54) has been obtained by methods such as multilocus enzyme electrophoresis and multilocus sequence typing (15) of housekeeping genes. The second objective is the tracing of strains causing outbreaks of disease in hospitals or a local community. Multilocus enzyme electrophoresis and multilocus sequence typing are not suitable for such investigations (23), so methods with increased discriminatory power are used. These include whole-genome analysis by pulsed-field gel electrophoresis (20), arbitrarily primed PCR, randomly amplified polymorphic DNA, amplified fragment length polymorphism analysis (34), enterobacterial repetitive intergenic consensus PCR, repetitive element sequence-based PCR (20), insertion sequence typing (37, 50), spacer oligonucleotide typing (spoligotyping) (14, 21), ribotyping (9), and study of microsatellites (35). These methods are highly discriminatory and rely on uncharacterized genomic differences between isolates of bacterial species. However, the procedures are laborious, may require large amounts of purified DNA (for pulsed-field gel electrophoresis), are somewhat subjective (arbitrarily primed PCR, ribotyping), and produce results that may not always be reproducible and are difficult to share between laboratories.

Another drawback of the methods listed above is the empirical choice of target sequences. Over recent years, full-genome sequencing has been performed on an increasing number of bacteria. This has provided useful information for taxonomic, evolutionary, and phylogenic purposes, but although it is the most detailed form of genotyping, full-genome sequencing is clearly not adapted to strain typing. The availability of full-genome sequences, however, enables the rational selection of target sequences in molecular studies. We speculated that the areas of the genome containing the most-variable sequences in closely related bacterial species would also be the most variable among strains of the species and hence the most useful for differentiation of strains. In plants, it has been demonstrated that noncoding sequences were superior to genes for the phylogenetic and genotypic classification of species (7, 10, 17, 18, 56, 61). We speculated that noncoding sequences, which should not be subject to selection pressure, would be appropriate for typing bacterial strains.

To test our hypotheses, we compared the genome sequences of two closely related bacteria and attempted to develop a rational method for selecting sequences suitable for strain typing. We used Rickettsia conorii in our study for several reasons, including (i) the availability and colinearity of the R. conorii and Rickettsia prowazekii genome sequences, which facilitate comparisons of various types of sequences, (ii) the absence of effective phenotypic (43) or genotypic (47-49, 51) typing methods at the strain level, and (iii) the availability in our laboratory of a large collection of R. conorii isolates. Because our purpose was not a taxonomic study of rickettsiae closely related to R. conorii but rather the development of a genotyping tool at the strain level for R. conorii sensu stricto (type strain, Malish), we did not include in our study the Astrakhan fever rickettsia, Israeli spotted fever rickettsia, and Indian tick typhus rickettsia, whose taxonomic status is uncertain and which can easily be differentiated from each other and from R. conorii sensu stricto on the basis of specific nucleotide substitutions within the ompA, ompB, and sca4 nucleotide sequences (45, 48, 51).

MATERIALS AND METHODS

Selection of target sequences.

We aligned the genome sequences of R. conorii (GenBank accession number NC_003103) and R. prowazekii (NC_000963) by using BLAST (1) and identified conserved or degraded fragments within both coding and noncoding sequences. To test whether noncoding sequences would be better candidates for R. conorii strain genotyping than coding sequences, we studied four conserved genes present in the genomes of both R. conorii and R. prowazekii (the 16S ribosomal DNA [rDNA], gltA, ompB, and sca4 [also referred to as gene D] genes [Table 1; Fig. 1] [45, 47-49, 51]); ompA, a coding gene present in R. conorii and used for species identification of tick-borne rickettsiae, which is present as a remnant in R. prowazekii (Table 1; Fig. 1) (45); 23 split genes in R. conorii which had statistical coding potential but were fragmented compared to their orthologs in R. prowazekii (Table 1; Fig. 1); 5 remnant genes in R. conorii which were identified as genomic regions exhibiting significant sequence similarity with bona fide genes in R. prowazekii but were too degraded to be identified as bona fide genes by gene-finding programs; 25 variable intergenic spacers, selected from 150- to 500-bp sequences separating 2 genes consecutive in both genomes, which had BLASTN scores of <75 between the two genomes; and 27 conserved intergenic spacers with BLASTN scores of >75 (Fig. 2).

TABLE 1.

Sequences of primers, amplicon sizes, annealing temperatures, and accession numbers used in this study

DNA target namea	Forward primer (5′-3′)	Reverse primer (5′-3′)	Amplicon size (bp)	Annealing temp (°C)	GenBank accession no.
Genes conserved in both genomes
16S rDNA	fD1, SFG3, SFG4b	Rp2, SFG2, SFG5, SFG6b	1,400	See reference 47	AF541999
gltA	CS1d, CS535d, Rp877pc	CS428r, CS890r, Rp1258nc	1,047	See reference 49	U59730
ompB	120M59, 120-607, 120-1378, 120AA2235, 120-2788, 120-3462, 120-4232d	120-807, 120-1497, 120-2399, 120-2988, 120-3599, 120-4346, 120-4879d	4,682	See reference 48	AF123721
sca4 (gene D)	D1f, D767f, D1219f, D1738f, D2338fe	D928r, D1390r, D1876r, D2482r, D3069re	2,725	See reference 51	AF163008
Genes degraded in one of the two genomes
ompA	190-70f	190-701f	632	52	U43806 (Malish), U45244 (Moroccan), U43798 (M1)
Remnant of Rp352	GCAAAACGGTTGACATTTGA	CTCAAGACAAAATGGGGAAAA	221	50	AY515514
Remnant of Rp543	ATTTTGTTAAAACATATGTAGCGGTAT	TTCAATAGACCTCTTCCCAAGC	250	50	AY515515
Remnant of Rp550	CGGCATTAGATTTTCTGCTTG	CCATGTCATTCTTGCTTTCG	230	50	AY515516
Remnant of Rp723	CTCTACTAAAGCAGATGTCGAAGG	AACGAGCAATACACCTGCAC	250	50	AY515517
Remnant of Rp820	AACCGATGCCTTTTGATTTG	CAACTGCTACATGCCCTGAA	288	50	AY515518
Split Rc148	TCTTTTTGAAGATTGGTGGAAAA	AAGGTTGATGCAAGGGACTTT	166	52	AY518488
Split Rc149	TTTGCAGGTACTGGCTTTATTTC	CCCTGAAAACCGTAACGAAT	159	52	AY518489
Split Rc215	CAAACCGGCTTGAATGATTT	AAGTATCGGCACTTCACATGC	174	52	AY518490
Split Rc217	ATGGAAGCGAGGAGAACCTA	TGCAGAAATATTATCGGTAAAAGC	105	52	AY518491
Split Rc269	CACAGGGGGCTTAAGTAGGTC	ACGGCAAACCAAATACTGTAAA	150	52	AY518492
Split Rc630	GGCATCATTATGCGGGAACAT	CCTCCTACAAGACCGACACC	106	52	AY518493
Split Rc653	AAGCATGCCGATGATTTACC	TGCAATTTCTTGAAGGCTTTT	156	52	AY518494
Split Rc654	TGCAAATCAAGGAGTAATGGTG	TTGCGTTATGCTTTGTATATTCG	157	52	AY518495
Split Rc655	GTCAAAACCGGAAATTGCAC	AAATGAAGGTTCGTTAAAAGCA	103	52	AY518496
Split Rc704	GAGACGATGGAATAGGAGTACTGAA	TTTTCTGCCCCAAATTTTTCC	157	52	AY518497
Split Rc721	TGAGACTCAAAGCCCTTATTTTTC	GGTGTTATTTCTTTATATCGCCAGT	150	52	AY518498
Split Rc776	GGAGGAGCTAAGGGAGCAAT	AAATTTCTTCAAATCGCTCAGG	160	52	AY518499
Split Rc777	TGGCTGATATAAAGGTAAGAAAACA	TCCTGCAATGTTTTTGTTTGA	156	52	AY518500
Split Rc837	GCTTGTGGTCTTGGTGTGG	AGCAAACATTTCGGTAACACC	150	52	AY518501
Split Rc838	GCCGAAAATGACGGTAAAAA	AGTGGAGCATGGTTTGCTGT	156	52	AY518502
Split Rc839	TTGGTTACAGCTGGAAATATGG	CACAAAGATATTTTACAAGAAAGTCA	187	52	AY518503
Split Rc840	GAGAAGCAGTTAATTGGACGTAAA	CAAGAAGCGCATCAAGATCA	150	52	AY518504
Split Rc925	ATTCCGGCATGGATGCTT	CCAAAGCACAAACCATAGAAAA	155	52	AY518505
Split Rc1042	GGTAGGGGCAAACAAAAGCTG	TGCCTTAAGTTTGAGTTGCTTGAA	196	52	AY518506
Split Rc1100	TGCTGTTTCTAGCTTAATGTG	CCTGTTTCTTTTTCCAACTTTTG	159	52	AY518507
Split Rc1101	TGCAGCTCAGGTTATTCATCA	TTAAAAGGTATGGTGCAAATAGAA	160	52	AY518508
Split Rc1102	AGGATTTAATTGGCGTCTTGC	TTTTTGTCGATATTTGCTTTTTCA	150	52	AY518509
Split Rc1137	CGCTTTCGTGAGAATGACAA	TCACCGTAGCTTCTCAAAGTTTC	161	52	AY518510
Intergenic spacers
nusG-rplK	CAGTTGCAATATTGGTAAAGCA	CAGCAGCTGGAATTATCAAGTT	270	54	AY345058
rpoB-rpoC	CAGGCATTCCTGAATCATTT	TCCGTAAAAATTTACTACGCTCA	257	54	AY345057
yqiX-gatB	CTGCGGCAGTACCGACTATT	ATCCGACGCTTGTGAATCAG	258	54	AY345059
rrf-pyrH	GAGCTTTCTCCATCTTTTCTTG	AAAGGGGAATATACGACAATTGAG	238	54	AY345060
RC0241-RC0242	AGCTCAAATTGTGGTGTTTCC	GGGATCCCTATTACAGCAAAA	343	54	AY345061
rne-coxW	CGGAAAAGAATGCAGAGTCTTG	CCATTTTGTAATTAAACTTTTCTGC	244	54	AY345062
RC0432-hsIV	TGTGTGGAGTTAATGTATATTGCGAT	CGAGACTTGTCCATCTGCTG	283	54	AY345063
asmA-rimM	TTAAAGGAATAAAGAAAGGAATAACAA	TTTTCAAATCAAACAACCGAAT	281	54	AY345064
murG-RC0563	GAAGAAAAGAAGGGCATAAGCTA	CAAGCTGAAAGTAAAAACATTCC	293	54	AY345065
dnaN-RC0584	TCGTCATGCCTGTTAAGGTG	TTGGATAATCACCCGCTAAGA	354	54	AY518297
lig-tgt	TTTTTGTGCTTCCTCTTCAGAT	CCAAAATCTCATGAGCCGTA	285	54	AY345066
rpsA-cmk	GCTGCAAGTTGTGGAACAAA	TTACCGGCTTCAGAGATGCTTT	254	54	AY345067
Rho-RC0760	CGGTQTTGTTAAGTTCTGCTGTG	TGCATGCCATTACTTATTACAAATG	385	54	AY345068
folC-bioY	AGGTCGGCACCGGAAAAT	TACGGCGGCGTATTACCTT	269	54	AY354069
tRNALeu-mgtE	AGCATTGAGGGTGCTGTTCT	TTCAGCAAATTGATCGTGATG	267	54	AY345072
pth-rplY	TTCCTGGATTACCAAGACCAA	GAAGCTGAAGGGGAACAACA	322	54	AY518299
ntrY-rpsU	AGCTGCTGTTGCTAAAGTAAAAA	CAAGAAGCAGCAAGAAGACAGA	363	54	AY345070
cspA-tRNALeu	CGCCATTGTCCTGTTCAATA	TCCGTTATGTCTACCATTCCA	375	54	AY345073
tmk-proP4	TTCCCCTCCCTCAAATGTAA	CGGAGCAAGAAACCCATAAA	258	54	AY354074
msbA2-RC1074	TCGAAATATTTGCAGAGAGCAG	TGAGCTCGCGAAAGTTAGAA	302	54	AY518298
proP5-RC1171	TGCGTGATTTTGTTTGTTTCA	GCACGTAAAATGGGAAAGTGA	261	54	AY345075
dksA-xerC	TCCCATAGGTAATTTAGGTGTTTC	TACTACCGCATATCCAATTAAAAA	416	54	See Fig. 4
serS-virB4	CGGATGTCTTGATAAATTACATGG	TCAAATTTTCGTAAACCACTAAACA	344	54	AY345076
tRNAIle-pal	GCGTGCTCTAACCAACTGAG	GAAGAAGCTTTTGCCTATAATCG	307	54	AY345079
pbpA2-RC0856	AGGTTTCCATTTTTCCCAAA	CGAGTAGAGTGZZGGZTZCTCGATG	343	54	AY518300
tRNAPhe-nifR3	TTGAACCAACGACACAAGGA	CCGTAACACCTGACATTGGA	251	54	AY345080
spo0J-abcT1	AAAGATTTGGAAGAATTAGACTTGAT	TTTGCTTAAACCAACCATTTCA	259	54	AY345081
RC0098-dcd	CCGATGCAAGGCAAATAATA	CGCAAAGGGCCTTATCATAC	288	54	AY345083
RC0102-RC0103	GCGATAAGCGATTTATTAGGC	GAAAGCCTAAAGCCTCCACA	240	54	AY345084
tRNAfMet-RC0138	GGTCGTTGGTTCAAATCCAG	AAGTCGTCATTGCGAGAAGG	265	54	AY521231
23S rRNA-5S rRNA	GTTGATAGGTCGGGTGTGGA	GGGATCGTGTGTTTCACTCA	200	54	AY345100
mppA-purC	GCAATTATCGGTCCGAATG	TTTCATTTATTTGTCTCAAAATTCA	160	54	AY345089 (type A), AY345087 (type B), AY345086 (type C), AY345085 (type D), AY465118 (type E)
tRNAGly-tRNATyr	AGCTTGGAAGGCTGGAACTC	ATCCTTCTCCCTCCACCACT	148	54	AY345097 (type A), AY345099 (type B)
rpmE-tRNAfMet	TTCCGGAAATGTAGTAAATCAATC	TCAGGTTATGAGCCTGACGA	144	54	AY345091 (type A), AY345092 (type B)
fabZ-lpxD	TGTTAGGATCGATTTTAAGTACTCTATCT	TGGATTGGCATAGACAATCTATTA	195	54	AY518302
fusA-tRNATrp	TGATCAAGTGCCGAGTCAAG	GCGCTCTACCAATTGAGCTAC	149	54	AY518303
RC0669-RC0670	TTTAATACCGTTAAACTTATCCAAGTG	TGTTCAACGCCATCATCTTC	295	54	AY518481
RC0280-23S RNA	CAAAAAGCCGACAAAGCCTA	CCTTCATCGCCTTCTAGTGC	258	54	AY518484
acrD-hupA	GGGCGTTTAATACAAATTTTAGACA	CAATTCTCCTTTGATAGGTTAATATGT	388	54	AY518473
RC1137-tlc-5	CGGGATAACGCCGAGTAATA	ATGCCGCTCTGAATTTGTTT	264	54	AY518475
pal-RC1201	TGCAAGCACACATAATGCAA	TCAAAATCGATTCCTCTTTTCC	216	54	AY518472
lgt-RC073	CCTCCGACTATTATGCCTATAACG	ATGACATTTCCTAATATCAATCCAA	244	54	AY518479
udg-RC1213	AATCCCACATATCCGCTACC	AGCCAAAGATAATGAAATCAGAA	248	54	AY518483
secB-czcR	ATGCAGGATTCCAGCCTTTA	GGCTCGCCTTCAATTAACAA	224	54	AY518486
RC0230-RC0231	TGCACCCGCCTAAAACTAAC	ATGGTCGGCCGTAGAAAAA	232	54	AY518480
groES-RC0970	CTTGCATCGGCTTTTCTTTT	AGCTTTGAGCTGATGGGCTA	215	54	AY518487
secA-prsA	GCAGGTTCAAGCGAGTTAATTT	AAAAGCAATACCGGAAAGCA	209	54	AY518485
RC1282-fdxA	CATGCCCTCAGCAAATGATA	GGTTCTGTGAAGATTGCTAATTGA	289	54	AY518482
RC0604-RC0605	AAAGGCAATAACGGCAAAAA	AGCTCGCCAGTTCATTCATC	353	54	AY518474
RC0409-trmU	AACCTTGACGTGCATATTCTAAA	GCCTGACATTGCGACAACTA	270	54	AY518477
RC0272-gyrA	AACAAGAATAGAGCAGCGTTCA	TTTCATCTCATCTTCGATATTTACC	368	54	AY518478
RC1027-xthA2	GGTATGTAAATGAGCCTTATCAATACT	TCAGTAGTATAAGTAGCTCCTGCTGTC	351	54	AY518476

^aIntergenic spacer designations consist of the name of the 5′ ORF- the name of the 3′ ORF. ORFs encoding putative proteins of unknown function have designations beginning with RC and are numbered with reference to the R. conorii Malish (Seven) genome (GenBank accession no. NC_003103).

^bSee reference 47.

^cSee reference 49.

^dSee reference 48.

^eSee reference 51.

^fSee reference 45.

FIG. 1.

Genome fragments selected for comparison of variability between R. prowazekii and R. conorii. Open boxes, ORFs present in both genomes; wavy lines, remnant in one Rickettsia species from a gene present in the other species; trisected box, gene split in R. conorii but not in R. prowazekii; solid lines between boxes, intergenic spacers exhibiting low interspecies variability; dashed lines between boxes, intergenic spacers with high interspecies variability.

FIG. 2.

Degree of intergenic spacer variability between R. conorii and R. prowazekii genomes as estimated by BLASTN scores. The most variable spacers exhibited low scores, whereas the most conserved spacers had high scores. White diamonds represent the spacers amplified and sequenced in our study that were not discriminatory among R. conorii strains; black diamonds represent the discriminatory spacers. Dashed horizontal line, BLASTN score of 75.

Rickettsial strains.

The 38 R. conorii strains studied were from the collection of our laboratory and are listed in Table 2. These strains had been isolated from ticks or clinical specimens from France, Spain, Portugal, Croatia, Russia, Turkey, Algeria, Tunisia, Morocco, Kenya, and Zimbabwe and had been identified as R. conorii by use of partial ompA sequencing (45). Sequences obtained from these 38 strains were also compared to the genome of R. conorii strain Malish (Seven) (accession number NC_003103). By comparison with R. conorii strain Malish (Seven) used for genome sequencing in 1999 (41), the R. conorii strain Malish (Seven) used in this study was the same strain but had been passaged in cell culture 60 times over the last 5 years. Overall, we compared the sequences from 39 R. conorii strains, including R. conorii strain Malish (Seven) with 2 different passage histories. We focused our study on strains of R. conorii sensu stricto.

TABLE 2.

R. conorii strains included in our study

Straina	Origin	Geographical origin	Reference	Supplier or source
Moroccan, ATCC VR-141T	Unknown	Morocco	6	ATCCb
Malish (Seven), ATCC VR-613Tc	Human	South Africa	Geard	ATCC
M1	Rhipicephalus sanguineus	Georgia, former Soviet Union	22	Gamaleya Institute, Moscow, Russia
Zim1	Haemaphysalis leachi	Zimbabwe	30	P. J. Kelly
ZimA	Rhipicephalus simus	Zimbabwe	30	P. J. Kelly
Kenya	Haemaphysalis leachi	Kenya	30	G. Dasch
Spain96	Human	Spain	8	N. Cardenosa
16-B	Human	Spain	8	N. Cardenosa
SV9	Human	Spain		N. Cardenosa
Portugal4S	Human	Portugal		F. Bacellar
Portugal454	Human	Portugal		F. Bacellar
Portugal821	Human	Portugal		F. Bacellar
URRCFrance1	Human	France
URRCFrance2	Human	France
URRCSpain3	Human	Spain
URRCFranceFEe4	Human	France
URRCFranceFEe5	Human	France
URRCFranceFEe6	Human	France
URRCFranceE7	Human	France
URRCFranceFEe8	Human	France
URRCFranceF9	Human	France
URRCFrance10	Human	France
URRCFranceFE11	Human	France
URRCFranceFE17	Human	France
URRCAlgeria18	Human	Algeria
URRCFranceFEe25	Human	France
URRCTunisia28	Human	Tunisia
URRCCroatia29	Human	Croatia
URRCFranceFEe31	Human	France
URRCFranceFEe32	Human	France
URRCFranceTick46	Rhipicephalus sanguineus	France
URRCFranceFEe48	Human	France
URRCFranceFEe49	Human	France
URRCFranceFEe53	Human	France
URRCFranceFEe57	Human	France
URRCTurkey58	Human	Turkey
URRCTurkey59	Human	Turkey
URRCTurkey61	Human	Turkey

^aStrain designations include the country in which the patient was infected or from which the tick was collected.

^bATCC, American Type Culture Collection.

^cStrain Malish (Seven) used for PCR amplification and sequencing in this study was the same as that used for genome sequencing in 1999 (GenBank accession no. NC_003103) but had undergone 60 cell culture passages since that date.

^dJ. H. Gear, personal communication.

Rickettsiae were propagated at 32°C on Vero cell (ATCC CRL-1587) monolayers in Eagle's minimal essential medium (Seromed, Berlin, Germany) supplemented with 4% fetal bovine serum (Seromed) and 2 mM glutamine. When cells stained with Gimenez stain were heavily infected (3 to 5 days), the cultures were harvested, centrifuged (12,000 × g for 10 min), resuspended in Eagle's minimal essential medium, and stored at −70°C until further processing.

PCR amplification and sequencing.

The primers used to amplify the fragments described were obtained from Eurogentec (Seraing, Belgium) and are listed in Table 1. Their specificity was predicted by using the BLAST software (1). Genomic DNA was extracted from the rickettsial cultures by using the QIAamp tissue kit (QIAGEN, Hilden, Germany) according to the manufacturer's instructions. PCRs were carried out in a PTC-200 automated thermal cycler (MJ Research, Waltham, Mass.). Two microliters of the DNA preparation was amplified in a 50-μl reaction mixture containing 50 pM each primer, 200 μM (each) dATP, dCTP, dGTP, and dTTP (Invitrogen, Gaithersburg, Md.), 1 U of eLONGase polymerase (Invitrogen), 2 μl of eLONGase buffer A, and 8 μl of eLONGase buffer B. The following conditions were used for amplification: an initial 3 min of denaturation at 94°C was followed by 40 cycles of denaturation for 30 s at 94°C, annealing for 30 s at various temperatures given in Table 1, and extension for 1 min at 68°C. Amplification was completed by holding the reaction mixture for 3 min at 68°C to allow complete extension of the PCR products. PCR products were purified by using a QIAquick Spin PCR purification kit (QIAGEN) as described by the manufacturer. Sequencing reactions were carried out by using the dRhodamine Terminator cycle sequencing ready reaction kit with Amplitaq polymerase FS (Perkin-Elmer, Coignieres, France) as described by the manufacturer. For all PCR products, sequences from both DNA strands were determined twice. Sequencing products were resolved by using an ABI 3100 automated sequencer (Perkin-Elmer). Sequence analysis was performed by using the ABI Prism DNA sequencing analysis software package (version 3.0; Perkin-Elmer). Sterile water was used as a negative control in each assay.

Sequence analysis.

In order to estimate “in silico” (by computer simulation) the relative variability of coding and intergenic sequences between the two genomes, we selected 102 pairs of putative orthologous intergenic sequences and 770 pairs of orthologous coding sequences that were reciprocal best hits with a significant BLASTN E-value of <0.001 and exhibited small size differences (<20%). Then we calculated the mean (± standard deviation [SD]) nucleotide sequence similarity for noncoding and coding sequences.

DNA sequences obtained in our study and those from the genome of R. conorii strain Malish (accession no. NC_003103) were aligned by using CLUSTALW software, version 1.81 (1). For split and remnant genes, for which the lengths of the sequences compared could be quite different, we considered consecutive gaps as a single mismatch. Percentages of similarity among sequences were determined by using the MEGA 2.1 software package (31). Phylogenetic relationships among R. conorii strains were inferred from both the intergenic spacer sequences and the multigene sequences by using the MEGA 2.1 software package (31). Distance matrices were determined under the assumptions of Kimura by using complete deletion analysis and were used to infer dendrograms by the unweighted pair group method with arithmetic means (UPGMA) available in the MEGA 2.1 software package (31).

Statistical analysis.

Student's t test was used to compare the nucleotide sequence similarity means of noncoding and coding sequences and of coding genes, split genes, remnant genes, and intergenic spacers. Multispacer typing (MST) and multigene typing were compared by using the χ² test. STATA software (version 7.0; Stata Corporation, College Station, Tex.) was used for statistical analysis.

Nucleotide sequence accession numbers.

The sequences reported in this paper have been deposited in GenBank (accession no. AY515514 to AY515518, AY518488 to AY518510, AY345057 to AY345070, AY345072 to AY345076, AY345079 to AY345081, AY345083, AY345084, AY518297 to AY518300, AY521231, AY345100, AY345085 to AY345087, AY345089, AY465118, AY345091, AY345092, AY345097, AY345099, AY518302, AY518303, AY518472 to AY518487, AY428738 to AY428750, AY462116, and AY497559).

RESULTS

In comparing the R. conorii and R. prowazekii genome sequences, we observed that the mean nucleotide sequence similarity (± SD) of the 25 variable spacers was 68.3% ± 7.8%. This was significantly lower than those of the 4 conserved coding genes (89.3% ± 7.8%; P < 10⁻²), the 23 split genes (83.1% ± 4.6%; P < 10⁻²), the 5 remnant genes (84.6% ± 7.5%; P < 10⁻²), and the conserved spacers (84.9% ± 2.3%; P < 10⁻²) but was similar to that of ompA (66.3%).

In the 38 strains of R. conorii for which we determined sequences in this study, the sequences of all 4 conserved coding genes, 23 split genes, and 5 remnant genes were identical to those of the genome of R. conorii strain Malish (Seven) (Table 1). Analysis of ompA sequences enabled us to identify three genotypes: one comprising 36 strains which had 100% similarity with the genome of R. conorii strain Malish (Seven), one containing R. conorii strain Moroccan, and one consisting of R. conorii strain M1 (Table 1). Overall, by using multigene typing with coding genes in R. conorii (16S rDNA, gltA, ompB, sca4, and ompA), we were able to classify the 39 strains of R. conorii we compared into three genotypes.

PCR amplification of the 52 intergenic spacers we studied in the 38 test strains of R. conorii yielded product sizes consistent with those of the genome of R. conorii strain Malish (Seven) (Table 1) except for the dksA-xerC spacer, for which the amplicon length ranged from 100 to 549 bp in different strains of R. conorii (Fig. 3). Only 4 spacers, all of which belonged to the 25 variable spacers (Fig. 2), had nucleotide differences in the R. conorii strains we studied: dksA-xerC, mppA-purC, rpmE-tRNA^fMet, and tRNA^Gly-tRNA^Tyr (Table 3). Seven nucleotide differences within the mppA-purC spacer enabled R. conorii strains to be classified into five genotypes (Table 3). By using differences at two positions in the rpmE-tRNA^fMet spacer, strains could be classified into two genotypes (Table 3). The tRNA^Gly-tRNA^Tyr spacer enabled the identification of two genotypes based on three nucleotide mutations. The dksA-xerC spacer was found to contain variable numbers of 63- to 102-bp repeat units designated R1 to R5, depending on the strain (Fig. 3 and 4). The percentage of nucleotide sequence similarity between repeats ranged from 50% between R1 and R4 to 98% between R1 and R2. Their G+C contents ranged from 11.8% for R5 to 17.5% for R3 (Fig. 3). Differences observed in the dksA-xerC spacers enabled us to classify the R. conorii strains we studied into 15 genotypes (Table 3). The number, type, and arrangement of repeat units for all strains tested are detailed in Fig. 4. The two batches of R. conorii strain Malish (Seven) exhibited identical spacer sequences. By combining the results obtained from analysis of the four variable spacers, we were able to identify 27 genotypes of R. conorii (Table 3). However, identification of these 27 genotypes was obtained by combining the results from the dksA-xerC, mppA-purC, and rpmE-tRNA^fMet spacers only. The tRNA^Gly-tRNA^Tyr-based typing did not provide any additional genotype. Therefore, this spacer was removed from the MST analysis. Twenty-three of the 27 genotypes contained only a single strain, while the remainder contained four strains (2 genotypes), three strains (2 genotypes), or two strains (1 genotype). MST identified a significantly greater number of genotypes than multigene typing (27 of 39 versus 3 of 39; P < 10⁻²).

FIG. 3.

Repeated sequences making up the dksA-xerC intergenic spacers of R. conorii strains.

TABLE 3.

Results of MST for R. conorii strains

Straina	Genotype derived from:
	dksA-xerC	mppA-purC	rpmE-tRNAfMet	MSTb
URRCCroatia29	A	B	B	1
16B	B	B	A	2
M1	B	B	B	3
Spain96	B	B	B	3
URRCTunisia28	B	B	B	3
URRCFranceFEe31	B	B	B	3
Portugal4S	B	D	A	4
URRCFranceFEe8	B	D	B	5
URRCFranceTick46	B	D	B	5
URRCFranceFE11	B	A	B	6
URRCAlgeria18	B	A	B	6
URRCFranceFEe25	B	A	B	6
URRCFranceFE17	C	A	B	7
URRCFranceFEe48	C	E	B	8
URRCFranceFEe53	C	A	B	9
URRCTurkey58	C	B	A	10
URRCFrance1	D	A	B	11
URRCFranceFEe5	D	A	B	11
URRCFranceF9	D	A	B	11
URRCFranceFEe32	D	A	B	11
URRCFranceFEe4	D	B	B	12
URRCFranceFEe57	D	A	A	13
ZimA	E	B	B	14
Zim1	E	C	B	15
URRCFranceFEe49	F	A	B	16
Malish (Seven)*	G	B	B	17
Malish (Seven)¶	G	B	B	17
Kenya	H	B	B	18
URRCSpain3	I	A	B	19
Portugal1454	I	B	A	20
SV9	I	B	B	21
URRCFranceFEe6	J	A	B	22
URRCFranceFE7	J	A	B	22
URRCFrance10	J	A	B	22
URRCFrance2	K	A	B	23
Portugal82I	L	B	A	24
Moroccan	M	B	B	25
URRCTurkey59	N	B	A	26
URRCTurkey61	O	B	A	27

^a*, strain Malish (Seven) used for genome sequencing; ¶, strain Malish (Seven) with 60 more culture passages.

^bMST uses combined results from the three spacers.

FIG. 4.

Number and repartition of VNTRs in the dksA-xerC intergenic spacers of 39 R. conorii strains. *, strain Malish (Seven) used for genome sequencing; ¶, strain Malish (Seven) with 60 more culture passages.

By comparing coding and intergenic sequences between the two genomes in silico, we found that the 102 intergenic sequences studied had a mean nucleotide sequence similarity (± SD) of 79.4% ± 13.5%, significantly lower than that of the 770 open reading frames (ORFs) (89.1% ± 3.9%; P < 10⁻²).

When sequences from the dksA-xerC, mppA-purC, and rpmE-tRNA^fMet variable spacers were concatenated, three clusters could be differentiated. One contained all 5 sub-Saharan African strains; another was composed of 16 strains, mostly from France; and the third cluster included 13 strains from various geographical locations, including strains Moroccan and M1 (Fig. 5). Five strains of R. conorii did not group within any of the clusters.

FIG. 5.

Unrooted tree showing the phylogenetic relationships among the 39 R. conorii strains studied, inferred from sequence analysis of the combination of the dksA-xerC, mppA-purC, and rpmE-tRNA^fMet intergenic spacers using the UPGMA method. The geographic origins of strains are given either in their names or in parentheses. *, strain Malish (Seven) used for genome sequencing; ¶, strain Malish (Seven) with 60 more culture passages.

DISCUSSION

We tested our hypothesis that the most suitable sequences for genotyping of bacterial strains are those that are found to be most variable when the genomes of two closely related bacteria are aligned. Using R. conorii and R. prowazekii, we found that the most variable sequences at the species level, variable intergenic spacers in this case, were also the most variable at the strain level and that suitable typing sequences were found in noncoding zones. Using a combination of sequences from three variable spacers in a multispacer tool, we identified 27 genotypes among the 39 strains of R. conorii we studied, including R. conorii strain Malish (Seven), for which we tested two batches with different passage histories.

Prior to our work, there was no genotyping method described for rickettsiae at the strain level. The development of a typing method for rickettsial strains has become crucial with the classification of R. prowazekii as a potential agent of bioterrorism. We used a rational technique rather than an empirical strategy to search for the most suitable genome fragments for this purpose. Comparison of the R. conorii and R. prowazekii genomes, which exhibit a high degree of colinearity, enabled us to compare the interspecies variability of coding genes, degraded genes, conserved intergenic spacers, and variable spacers. By in silico analysis, we found that variable intergenic sequences were more variable than coding genes, degraded genes, and conserved spacers (P < 10⁻² in all cases). It has been suggested that intergenic spacer sequences are an important source of genome plasticity because they do not undergo selection pressure (13). For the rickettsiae, it has been suggested that most of the intergenic sequences of R. prowazekii and R. conorii consist of decayed genes that are no longer active but have not yet been totally eliminated from the genome (42). To date, the intergenic spacer used most widely in work with bacteria has been the 16S-23S rDNA spacer. In many bacterial species (24, 25, 32, 36, 46, 55) this spacer has been shown to have great variability, not only in its sequence and length but also in the number of alleles per genome (33, 53). However, studying the 16S-23S rDNA spacer of rickettsiae is not possible, because the 16S rDNA gene is separated from the 23S and 5S rDNA genes, which are tightly linked together (5).

When we compared the overall variability of intergenic spacers and coding sequences between the genomes of R. prowazekii and R. conorii, we found that spacers were significantly more variable than genes (P < 10⁻²). We were surprised, however, to observe that there were both highly variable spacers and spacers that were as conserved as coding sequences. R. conorii and R. prowazekii are estimated to have diverged from a common ancestor 80 million years ago (42), and it would seem probable that sequences which were not under selection pressure during this period would have been more variable than coding sequences. For prokaryotes, variations in the conservation of spacers has been reported, but the role of conserved spacers is incompletely understood (39, 44). For eukaryotes, recent studies on comparative genomics have shown that in yeasts (29) and mammals (11) some intergenic spacers are highly conserved at the species level. Some of these spacers include regulatory motifs, but the function of many remains unidentified. The factors responsible for the heterogeneity of intergenic spacers in rickettsial genomes have yet to be determined. Nevertheless, our results emphasize the importance of comparing genomes in order to select variable sequences instead of targeting sequences presumed to be variable.

At the intraspecies level, we confirmed that the 25 variable spacers we studied were the best targets, with 4 being highly variable and a combination of 3 enabling us to identify 27 genotypes among the 39 strains of R. conorii we studied. As predicted by in silico analysis, conserved genes were highly conserved among the R. conorii strains in our experiments. Only one of the coding genes we studied exhibited interstrain variability. This was ompA, which encodes a high-molecular-weight, surface-exposed protein and is one of the most variable coding genes in the spotted-fever-group rickettsiae (16). The variability we found in the gene was limited, however, and enabled us to differentiate only three genotypes. ompA-based genotyping was not congruent with MST, which classified strain M1 in a genotype with other strains. However, because the phylogenetic analysis using MST sequences was congruent with the geographic origins of strains, MST may be more relevant than ompA for R. conorii strain typing. Even when all five coding genes were used in multigene typing, there was less variability than that found with MST (P < 10⁻²). One could expect that genes which have undergone degradation since the divergence of R. conorii and R. prowazekii would exhibit higher levels of sequence divergence than conserved genes (3, 4) and thus would be better candidates for strain typing. We found, however, that split genes and remnant genes in R. conorii were also highly conserved and were not suitable for genotyping at the strain level. These findings support our strategy of selecting intergenic sequences with the greatest interspecies variability as targets for strain typing.

One of the 34 human isolates studied, URRCFranceFEe48, was obtained from a patient with malignant boutonneuse fever from Marseilles. Although it is likely that host factors play a key role in the development of severe forms of Mediterranean spotted fever (12), the specific role of strain variation in R. conorii has not been evaluated. The demonstration that this strain was a specific genotype by MST highlights the usefulness of our genotyping method for isolates associated with particular clinical presentations. In addition, the phylogenetic classification inferred from all four spacers was consistent with the geographic distribution of strains (Fig. 5).

In order to estimate the effect of culture passage on MST variation, we compared the intergenic sequences of two batches of R. conorii strain Malish (Seven) with different passage histories. No difference in spacer sequence was found between the two batches. Therefore, MST may be valuable for tracing rickettsial isolates from a single source with a difference in culture history of at least 60 passages.

Among the four variable spacers, the dksA-xerC spacer was composed of 63- to 102-bp repeat units (Fig. 3). One salient feature of the R. conorii genome is the high density of repeated sequences (40). Six hundred fifty-six interspersed repeated sequences, named Rickettsia palindromic elements (RPE), have been identified in R. conorii and represent 3.2% of the genome (2, 41, 42). Such interspersed repeats are usually confined to the intergenic regions of bacterial genomes (59), but in R. conorii the repeated sequences are also present within protein-coding regions (41). The repeats we found in the dksA-xerC intergenic spacer were highly conserved and were present only in this locus. Repeats representing a single locus and showing interindividual length variability are designated VNTRs (variable number of tandem repeats) (59). Changes in the number of repeats in a given genetic locus are an important source of DNA variability in eukaryotes (58). Such markers are well-established molecular targets for pedigree analysis in humans (27) and have also been used for bacteria (28). The VNTRs we found within the dksA-xerC intergenic spacer of R. conorii had two peculiar features. First, they occurred within an intergenic spacer, whereas VNTRs usually appear to be mainly involved in implementing size variation in cell wall- or membrane-associated proteins. This may cause enhanced or diminished exposure of active protein domains on bacterial surfaces (59). In Rickettsia species, VNTRs are known to occur within the ompA gene (59), and their number and arrangement vary in Rickettsia species (19). Second, the G+C content of the dksA-xerC VNTRs was low in contrast with that of the GC-rich RPEs common in intergenic regions of spotted-fever-group rickettsiae (3). In view of the fact that the overall G+C content of Rickettsia species is low, the dksA-xerC VNTRs may be remnants of a decaying rickettsial gene and not imported elements.

Our study demonstrated that in silico identification of the most variable sequences between two closely related bacterial genomes enables the selection of target sequences for strain genotyping. For rickettsiae, we demonstrated that intergenic spacers, in particular those that showed the greatest variability between the genomes of two closely related species, are more suitable for strain genotyping than coding sequences and degraded genes. The combined use of variable spacer sequences, which we named multispacer typing, is significantly more discriminatory than multigene sequencing. The advantages of MST include high discrimination, reproducibility, simplicity of interpretation (because one technique is used rather than a combination of techniques), and ease of incorporation of the data generated into databases that are directly comparable and readily shared by laboratories via the Internet. This technique may be applied for tracking isolates obtained from a wide variety of sources, including isolates from a single strain with different passage histories, and may even be applied directly to clinical specimens. Moreover, this technique may be applicable to other bacteria, in particular those considered potential agents of bioterrorism.